soft3/audit/assumptions/2026-10-30.md

assumptions — receipt 2026-10-30

This is the first observation receipt of the assumptions registry. It holds the status, best attack and measurement dates of each seed entry. The statements, consumers and responses stay in the proposal, and they move to specs/assumptions.md once it lands. The receipt was written during review round 1 of soft3#398.

Checked against these origin revisions:

repository revision
cybernode main 52538a6
hemera main fd55f99 (campaign text as at 76dc941)
zheng master ecf2491 (ledger computed on hemera 0.3.1)
tru master e660d1e
simulation main 5483a3f (receipt of 2026-10-18)
eidos main c1e8993
openai/math 42907c9 (ETH preprint, 2026-10-26)
id status evidence measured
A-01, A-02 holds (theorem) zheng ledger: "proven"; ε per round (d + 1)/p³ = 2^-189; succinct IOP ≤ 2^-173.5 at m = 2^20; zk IOP 119/p³ = 2^-185.1 at log m = 10 ledger at ecf2491
A-03 holds (tested) conformance: 2,000 random blocks, conserved at every height (cybernode rs/src/chain/economy/vectors.rs:318–364); simulator: conservation every block, e.g. reorg-140, 4,400 block applications on 9 nodes 2026-10-18
A-04 holds (tested) simulator launch-pussy, launch-bostrom: equal records on every node; every run bit-reproducible on macOS arm64 and Linux x86_64 2026-10-18
A-05 holds (tested) daa.rs tests: floor division, conformance §4 rows, clamp to 1, saturation at the limit cybernode 52538a6
A-06 holds (tested) genesis verify over both genesis files (genesis receipts audit/genesis-2026-10-15.md, audit/genesis-dimension-v3-2026-10-16.md) 2026-10-16
A-07 holds (measured) bostrom power-law, ℓ1 to the 192-step reference after 48 steps: 7.5 × 10⁻⁸; observed rate ≈ 0.75 per step against the a-priori κ = 0.775; with the ground amendment, 0.9–1.3 × 10⁻¹⁰ (the rounding floor) tru audit §2.1, §6.3, 2026-10-17
A-08 red soft3#234 (draft): finding 1, a quiet-epoch relock is paid 10,333,427 against 3,757,610 for the first lock; finding 4, an equal +1/−1 hedge credits the +1 side 5,315,353 and the −1 side 0 2026-10-25
A-09 red soft3#234 finding 3: J(φ*₀) on bostrom is 9,886,742,015,786,730,065 and is recorded as 9,223,372,036,854,775,807 (i64::MAX) 2026-10-18
A-10 trusted, unproved six constants in eidos/rs/src/stdlib/mod.rs:114–158; three typed Prop as placeholders eidos c1e8993
B-01 holds on generic bounds campaign: 0.4 has no row below its target; 0.3 (not frozen) was 14.6–25.8 bits over the target on full preimage (estimate). The campaign is a draft and day 0 has not started 2026-10-16
B-02 holds on the inequalities 0.4: every Poseidon2 inequality holds (minimal R_P at R_F 12 is 14; 48 shipped); subspace margins +33 linear, +18 nonlinear. 0.3: CICO-1 degree 2^38.46 < p, reduced-round CICO-1 solved in 92 s (results log 1) 2026-10-16
B-03 holds the header crosses at least two permutations (campaign §3.2) 2026-10-16
C-01, C-02 named ROM: "hemera-as-RO conjectured" in the ledger; QROM: G5 open ledger ecf2491
C-03 proven at the shipped regime ledger WHIR row: Johnson-regime MCA (BCGM ePrint 2025/2051 Lemma 9.3) ledger ecf2491
C-04, C-05, C-06 conjectured ledger header: "conjectured rows remain … so the repair does not pass that gate yet" ledger ecf2491
D-00 context ETH preprint dated 17 Oct, revised 25 Oct, committed 26 Oct; its Lean scope is weaker by (log log n)² against log log n (lean/docs/378.md). arXiv:2610.06783 v1 2026-10-05 2026-10-30
E-01 red every phase-1 signature is secp256k1 (chain.md §2.7); no sunset for transaction authorization (privacy-frozen §1.9, §6 item 8) 2026-10-30
E-02 not in use rule set 2 is off —
F-01 holds; captures late joiners without checkpoints reorg-140: a 140-deep rewrite by a 60 % attacker is refused by every honest node; a late joiner at 490 follows the attacker's chain with checkpoints off, and not with them (reorg-140-checkpoints) 2026-10-18
F-02 holds to 1 h a 30-min eclipse heals; a 150-min eclipse by a 5 % miner (292 blocks) strands the victim; a 2-h partition needs an operator resync 2026-10-18
F-03 holds clock-skew: all checks pass 2026-10-18
F-04 red settle-panic: 4/4 nodes crash at settle(1) and crash again after restart; settle-slow: waits and resumes; bostrom settlement max 52,003 ms against D = 1 h 2026-10-18
F-05 holds, slowly ÷100: first block 6,929 s after the step (exponential, mean 3,000 s, P(> 1 h) = e^(−1.2)); longest gap 6,957 s; back to 30 s ± 20 % at 8,613 s. chain.md §1.5 and §8 say 66 min 2026-10-18
G-02 measured economy: a concentrator was paid 285,656 against 27,966 for honest diverse stakers 2026-10-18
G-01, G-03, G-04, G-05 choices in force chain-economy §0, §12 —
H-01 red both genesis files carry an all-zero checkpoint_key; no signer exists; a key changes both genesis particles and needs a re-bless before 2026-11-02 00:00 UTC (soft3#398, director's comment) 2026-10-18, 2026-10-27
H-02 holds doctrine —
H-03 pending the bounty terms and the private reporting channel are the owner's to open (launch.md, decision 2026-10-16) —

Next review: on any change to a consumer of a red entry, at hemera's campaign day 0, and at every release of openai/math that touches a watch field.

Homonyms

2026-10-30
the beat · 2026-10-30 the launch was split in two, and the first of the two got a pinned set of sources. 2026-10-30-launch-split|the launch split in two: pussy attempts 2026-11-05 with its holes named in public, bostrom launches by readiness, fathership runs in parallel. soft3 re-scoped its plan on…

Graph